台湾生产率增长与市场不完善

Journal of Productivity Analysis, 22, 5–27, 2004 # 2004 Kluwer Academic Publishers. Manufactured in The Netherlands. Market Imperfection and Productivity Growth—Alternative Estimates for Taiwan CHIA-HUNG SUN ecdchs@ccu.edu.tw Department of Economics, National Chung Cheng University, Chia-Yi, 621, Taiwan, R.O.C. Department of Economics, Research School of Pacific and Asian Studies, Australian National University, ACT 0200, Australia Abstract This study first explores why the shares of factor inputs have not been measured correctly and concludes that the earlier findings are biased due to the miscalculation of factor shares which have produced low estimated total factor productivity (TFP) growth in the East Asian countries. Second, three approaches are proposed to empirically illustrate the impact of capital and labor shares on the estimates of TFP growth. It is suggested that TFP growth in the East Asian economies will be understated if net indirect taxes and imperfect competition profit are ignored. Finally, by taking the net indirect taxes and imperfect competition profits into account, the result of this paper indicates that Taiwan’s economy has enjoyed an average annual TFP growth rate of 3.6% over the period 1980–1999. JEL Classification: O47, O53 Keywords: growth accounting, total factor productivity growth, Taiwan, market imperfection 1. Introduction The findings of Kim and Lau (1994), Krugman (1994), Young (1995) and Collins and Bosworth (1996) that total factor productivity (TFP) growth had little to do with the economic miracle achieved by four East Asian countries—Hong Kong, Korea, Singapore and Taiwan—have drawn considerable attention to the controversial debate. Although there is a consensus on the importance of TFP growth in the process of economic growth, the major concern of the debate is that different methods or assumptions used to calculate it have often led to different results. The role of TFP growth in East Asia is not only crucial for the future of the region but of particular importance for less developed countries because the successful experience can serve as a model for them to follow. Nonetheless, prevailing TFP studies fail to explain why there are so many different results, even when similar methodology is used, namely, growth accounting, for the same country or region (see, for example, Collins and Bosworth, 1996; Klenow and Rodriguez- Clare, 1997; Sarel, 1995; Young, 1992, 1995).1 No serious attempts have been made to resolve the issue to date. Therefore, this study endeavors to identify what caused the diverse results. Two alternative explanations are offered for the discrepancies. The first possibility relates to the calculation of inputs growth. Whether or not the adjustment of embodied quality improvement in capital and labor inputs has been made will generate different input growth rates. Even when the adjustment is carried out, results may differ due to varying approaches to the adjustment (see, for example, Young, 1995; Fuess and Van den Berg, 1996; Collins and Bosworth, 1996).2 Moreover, the inclusion of human capital in growth accounting by some economists seems more controversial, especially on the issue of how to derive human capital stock (Collins and Bosworth, 1996; Klenow and Rodriguez-Clare, 1997). Under these circumstances, it comes as no surprise that there are so many differing empirical results. The second possibility is that the estimation of capital and labor shares (or elasticities of output with respect to inputs) has been more crucial for the outcomes of TFP growth estimates. Hence, one of the objectives of this study is to argue that the miscalculation of factor shares is responsible for the low estimated TFP growth in previous studies.3 Since the growth rate of capital input usually exceeds that of labor input in most countries, the higher capital share implies the lower TFP growth.4 Consequently, TFP growth rates have most likely been underestimated in prior studies due to the use of incorrect estimates of capital share, for example, Kim and Lau (1994), Young (1995) and Collins and Bosworth (1996) for the East Asian countries. How could the miscalculation of factor shares occur in most previous TFP studies? Failure to consider the effect of net indirect taxes (¼ indirect taxes less subsidies) and market imperfection underestimates the share of the contribution of labor input to factor income, which in turn leads to an overestimation of capital share and an understatement of TFP growth.5 As subsidies in value added contribute to production but indirect taxes do not, the importance of allowing for net indirect taxes need not be emphasized.6 Not only does the extent of imperfect competition differ from sector to sector, few would disagree with the existence of market imperfection in Taiwan, predominantly, in the finance and utilities sectors. In fact, the reality of market imperfection has been empirically substantiated by Hall (1988).7 Temple (1997, p. 281) also notices that the results of Young (1995) are subject to the possible effect of imperfect competition. Hsieh (1999) states that with imperfect competition the TFP growth estimates of growth accounting will be biased. Put differently, the idea of growth accounting is built on two major assumptions comprising perfect competition and constant returns to scale. However, the existing data from the national accounts are not the case as described by the theoretical model. To ensure the validity of growth accounting, these issues must be seriously considered in the estimation of the factor shares. Meanwhile, Sarel (1995) implicitly expresses that the difficulty of estimating capital and labor shares may cause results to be fragile. Using a translog value added production function and the modified capital and labor shares, this study attempts to reinvestigate TFP growth in Taiwan over the 6 SUN period 1980–1999 except for the agriculture sector and government services. Although the study is confined to Taiwan because it is easier to examine the proposed hypotheses in greater depth for a single country, it is desirable to extend the coverage of comparative studies to other countries. The remainder of the paper is organized as follows. Section 2 provides an indication of market imperfection in the case of Taiwan. Evidence of net indirect taxes is presented in the Appendix. Section 3 demonstrates the empirical model and details variable adjustments and construc- tions. Section 4 explains the empirical results and makes comparisons with earlier studies. The summary and concluding remarks are made in Section 5. 2. Indication of Market Imperfection The key element that affects the calculation of factor shares is the adjustment (or removal) of imperfect competition profits. How can the extent of imperfect competition profits be measured? According to Taiwan’s national accounts, ‘‘operating surplus’’ includes rent, interest and profits. Alternatively, it can be split into two components as a result of market imperfection. One is the payment to capital input and the other is imperfect competition profits. Table 1 reveals that average operating surplus as a share of GDP is significant in a number of sectors in Taiwan, such as finance (0.662*0.672), utilities (0.413*0.532) and commerce (0.399*0.456) over the 1961–1999 period and mining sector (0.437) in the 1990s. One reason that may explain the relatively higher ratios of operating Table 1. The average operating surplus as a share of GDP in Taiwan. Sectors 1961–1970 1970–1980 1980–1990 1990–1999 1961–1999 Mining 0.089 0.102 0.114 0.437 0.186 Manufacturing 0.309 0.317 0.294 0.275 0.300 Utilities 0.529 0.532 0.518 0.413 0.498 Construction 0.103 0.180 0.209 0.216 0.177 Commerce 0.399 0.418 0.428 0.456 0.424 Transport 0.197 0.225 0.228 0.299 0.236 Finance 0.662 0.671 0.672 0.670 0.669 Business serv. — — 0.376 0.294 0.335 Community serv. 0.351 0.268 0.221 0.246 0.271 Economy 0.362 0.369 0.369 0.394 0.374 Notes: 1. The share for the finance sector during the period 1961–80 includes the business services sector but the business services sector is separated from the finance sector since 1981. Thus, the shares of 0.026 and 0.025 for the business services sector denote over the periods 1981–90 and 1981–99, respectively. 2. The GDP for nine major sectors means sectoral GDP. Source: The data of net indirect taxes are from ‘‘National Income in Taiwan Area of the Republic of China’’ published by Directorate-General Budget, Accounting and Statistics (DGBAS), the Republic of China, various issues. MARKET IMPERFECTION AND PRODUCTIVITY GROWTH 7 surplus to GDP is the existence of market imperfection, in particular, in the utilities and finance sectors. The former has long been monopolistic and the latter is regarded as being oligopolistic. The issue of market imperfection is not new as Young (1995, p. 648) states: ‘‘The absence of perfect competition, in the context of a constant returns to scale production function, could lead to mismeasurement of the elasticity of output with respect to each input, as factor shares need no longer reflect output elasticities. In particular, to the degree that monopoly profits are reflected in capital income, capital’s income share will tend to overstate the elasticity of output with respect to capital.’’ The consequence of such mismeasurement is that the labor share is understated resulting in the capital share being overestimated given the assumption of constant returns to scale. Therefore, low estimated TFP growth in prior TFP studies is most likely attributable to the mismeasurement of capital share. In general, the miscalculation of TFP growth becomes more significant if the incorrect capital share is applied to some specific sectors, for instance, the utilities, finance and manufacturing sectors. Owing to the existence of market power in these sectors, imperfect competition profits cannot be ignored.8 In terms of the manufacturing sector, profits usually have to be distributed to shareholders and re-invested to expand business and operation. As the majority of utilities in Taiwan are owned and run by the government; most of the profits have been counted as part of the central government’s income to secure a balanced budget. Analogous to the utilities sector, imperfect competition can be observed in the finance and insurance sector. The state- owned banks had dominated the local market for several decades prior to the passing of the New Banking Law in 1989, which thereafter deregulated privately owned commercial banks. Evidence of net indirect taxes is discussed in the Appendix 1. Despite the substantial work done on TFP studies, none have considered the adjustment for both net indirect taxes and imperfect competition profits.9 In terms of cross-country TFP studies, the World Bank (1993) studies 87 countries, Fischer (1993) presents TFP growth for 68 countries, Young (1994) estimates TFP growth rates for 118 countries, Collins and Bosworth (1996) perform estimations for 88 countries, and Klenow and Rodriguez-Clare’s (1997) study covers 98 countries. As for TFP studies on East Asia and the Asia-Pacific region, Kim and Lau (1994), Young (1995), Sarel (1995) and Hsieh (1999) present contradictory outcomes for the four East Asian economies, namely Hong Kong, Korea, Taiwan and Singapore. Besides the East Asian economies, Drysdale and Huang (1997), Chang and Luh (1999) and Fare et al. (2001) extend the coverage to the Asia-Pacific region and investigate the TFP progress of 17 APEC economies. Narrowing down to the TFP studies on Taiwan, Liang (1995), Fuess and Van den Berg (1996) and Hu and Chan (1999) employ a similar approach (growth accounting) but arrive at different results due to assorted adjustments for quality improvement in capital and labor inputs as well as different specifications for the production function.10 Regardless of Taiwan’s 8 SUN better performance in TFP among the East Asian economies, the results of Kim and Lau (1994), Young (1995) and Collins and Bosworth (1996) display a downward bias as a result of mismeasured factor shares. 3. Methodology, Variables Constructions and Adjustments Following Jorgenson, Gollop and Fraumeni (1987), and Young (1995), value added is specified as a translog function of capital and labor inputs: lnY ¼ a0 þ aK lnK þ aL lnLþ aT ?T þ 1 2 bKKðlnKÞ2 þ bKL lnK lnL þ bKT lnK ?T þ 1 2 bLLðlnLÞ2 þ bLT lnL ?T þ 1 2 bTT ?T 2; ð1Þ where Y, K, L and T denote value added, capital input, labor input, and time. Under the assumption of constant returns to scale, the parameters satisfy the following conditions: aK þ aL ¼ 1; bKK þ bKL ¼ bKT þ bLT ¼ bKL þ bLL ¼ 0: ð2Þ Because the data sets are only available at discrete points of time, say T and T � 1, the growth rate of output can be expressed as a first difference of lnYðTÞ and lnYðT � 1Þ: lnYðTÞ � lnYðT � 1Þ ¼ SK ½lnKðTÞ � lnKðT � 1Þ� þ SL½lnLðTÞ � lnLðT � 1Þ� þ TFPT ; ð3Þ where SK and SL represent the elasticities of output with respect to capital and labor inputs and Si ¼ ½SiðTÞ þ SiðT � 1Þ�=2, i ¼ K ;L and TFPT ¼ ½TFPðTÞþ TFPðT � 1Þ�=2. The expression of the average rate of technical change, TFPT , is also called as the translog index of the rate of total factor productivity growth, where TFPðTÞ and TFPðT � 1Þ denote the level of total factor productivity at time T and T � 1, respectively. The translog index is often referred to as the discrete version of Divisia index or the To¨rnqvist index. Under the assumption of perfect competition, the elasticity with respect to each input is equal to its share in total factor payments. Since the sum of capital and labor shares is unity, the capital share can be obtained by one less labor share.11 Because aggregate capital and labor inputs consist of a number of components such as machinery, transport equipment and buildings etc., aggregate capital and MARKET IMPERFECTION AND PRODUCTIVITY GROWTH 9 labor inputs are assumed to be the translog function of their components: lnK ¼ aK1 lnK1 þ aK2 lnK2 þ � � � þ aKM lnKM þ 1 2 bK11ðlnK1Þ2 þ bK12 lnK1 lnK2 þ � � � þ 1 2 bKMM lnðKMÞ2; ð4Þ lnL ¼ aL1 lnL1 þ aL2 lnL2 þ � � � þ aLN lnLN þ 1 2 bL11ðlnL1Þ2 þ bL12 lnL1 lnL2 þ � � � þ 1 2 bLNN lnðLNÞ2: ð5Þ Similarly, under the assumption of constant returns to scale, the parameters in equation (4) again satisfy the following conditions:12 X p aKp ¼ 1 and X p bKpq ¼ X q bKpq ¼ 0; p 6¼ q and p; q ¼ 1; 2; . . .M ð6Þ Thus, taking first difference of the equations (4) and (5) provides the growth rates of aggregate capital and labor inputs as weighted averages of the growth rates of their subinputs: lnKðTÞ � lnKðT � 1Þ ¼ X sKm½lnKmðTÞ � lnKmðT � 1Þ�; ð7Þ lnLðTÞ � lnLðT � 1Þ ¼ X sLn½lnLnðTÞ � lnLnðT � 1Þ�; ð8Þ where sij ¼ ½sijðTÞ þ sijðT � 1Þ�=2, i ¼ K ;L, j ¼ m; n, m ¼ 1; 2; . . .M and n ¼ 1; 2; . . .N. sij denotes the elasticity of each aggregate input with respect to each of its component subinputs, that is, assuming perfect competition, the share of each subinput in total payments to its aggregate factor. The expressions for the capital and labor input in equations (7) and (8) are considered as translog indices of capital and labor inputs. In fact, the indices adjust for quality improvement of aggregate capital and labor inputs. Jorgenson et al. (1987) illustrate the importance of disaggregating the inputs by quality levels; for example, labor input is classified by sex, age, education, employment status and occupation of employees. As can be seen from equation (8), the growth rate of aggregate labor input is a weighted average of the growth rates of subinputs, weights being the associated income shares, sLn. Hence, if the average education level rises over time, the procedure will capture the quality improvement of labor input by assigning a higher weight for category n because of the higher wage, wn. Finally, if TFP growth is interpreted as a shift in an aggregate production, the associated variables have to be measured as flows. Therefore, the flow of labor services is assumed to be proportional to total hours of work and the flow of capital 10 SUN services is proportional to the estimated capital stock, that is, LnðTÞ ¼ gLnHnðTÞ and KmðTÞ ¼ gkmCmðTÞ, with lnKðTÞ � lnKðT � 1Þ ¼ X sKm½lnCmðTÞ � lnCmðT � 1Þ�; ð9Þ lnLðTÞ � lnLðT � 1Þ ¼ X sLn½lnHnðTÞ � lnHnðT � 1Þ�; ð10Þ where Hn and Cm denote the total hours of work and estimated capital stock, respectively. 3.1. Capital Input The study applies the perpetual inventory approach to obtain the capital stock for each sector.13 Then, the investment series (gross fixed capital formation) is divided into five subinputs ðImÞ: residential buildings, non-residential buildings, construc- tions, transportation equipment, and machinery and other equipment. Land input is excluded due to lack of data. Under the assumption of smooth investment in the past, initial capital stock equals initial investment divided by the rate of depreciation ðdmÞ plus the growth rate ðgmÞ of the investment of entire period in each subinput, that is, Cmð0Þ ¼ Imð0Þ=ðdm þ gmÞ, where m ¼ 1; 2; . . . ; 5.14 Once the initial capital stock is derived, the estimation of capital stock ðCTÞ in each sector at time T follows the perpetual inventory method with geometric depreciation: CT ¼ X5 m¼1 ½Cm;T�1ð1� dmÞ þ Im;T�1�: ð11Þ 3.2. Labor Input In order to consider the quality improvement embodied in labor input, it is desirable to disaggregate labor input in each sector into several categories and compute the weighted growth rate of labor input by its share of the total payment.15 If the share of well-educated workers increases, quality improvement in labor input will be reflected in the translog labor index because of relatively higher wages paid to well- educated employees. Other adjustments of labor input growth are more or less ad hoc such as Fuess and Van den Berg (1996) and Collins and Bosworth (1996).16 Surprisingly, the adjustment of quality improvement in labor input did not generate a sizeable change in Taiwan as reported by Young (1995, Table VIII). The quality-adjusted annual growth rate of labor input merely increased by 0.3 percentage points at the economy level during the 1966–90 period as shown in Table 2. Eventually, the estimated TFP growth rate of the overall economy rose by 0.2 percentage points.17 For the manufacturing industry, other industry and services industry, the annual quality improvements in labor input are estimated to be 0.4, 0.3 and 0.2%, respectively. This study also adopts Young’s (1995) results to derive the MARKET IMPERFECTION AND PRODUCTIVITY GROWTH 11 adjustment in labor input.18 Therefore, the quality-adjusted growth rate of labor input is calculated by the sum of the growth rate of hours worked plus 0.3% labor quality improvement for the nine sectors and the overall economy. 3.3. Capital and Labor Shares Various estimates of capital and labor shares have been cited in the recent literature; for example, the capital and labor shares by Kim and Lau (1994) for Taiwan are 0.505 and 0.413 and they, respectively, become 0.257 and 0.743 in Young (1995). However, the factor shares estimates are not available in Liang (1995), Sarel (1995), Fuess and Van den Berg (1996), Hu and Chan (1999) and so forth, making it difficult to evaluate their empirical results. Alternatively, the shares of capital and labor could be chosen without explicit estimation, for instance, capital and labor shares being assumed to be 0.35 and 0.65 in Collins and Bosworth (1996). To derive the shares of capital and labor, it is often easier to begin with labor because there is more information on wages and employment. Due to lack of wage data for employers, own-account workers and family workers, the adjustment between the number of employees and employment is implemented by assuming that employers, own-account workers, and unpaid family workers earn the same wage as emp